Daniel D. Corkill, Kevin Q. Gallagher
The run-time performance of a blackboard-based application can be significantly improved by selecting an appropriate blackboard database representation. We present empirical validation of this statement by tuning the representation used in a large, blackboard-based AI application. Dramatic performance gains were obtained without changing my problem solving or control activities. The results underscore the importance of efficient blackboard database operations and the benefits of a flexible, instrumented blackboard development environment when tuning the blackboard representation. This investigation was facilitated by use of the Generic Blackboard Development system (GBB) to construct the application. GBB provides the flexibility to quickly change the database implementation without recoding. Similar performance tuning capabilities are available to any application written using GBB.